Caddy: Setting headers per file

[gavin19]

With NGINX I was able to set the header per filetypes in a folder, like

location /tuts (.*\.pdf) {
    add_header Content-disposition "attachment";
}

to ensure files like pdf/mp4 etc are always forced to download, regardless of browser.

I have something like

tuts.example.com {
    root /var/www/static/tuts
    browse
    header / Content-disposition "attachment"
    log ../logs/tuts.log
}

but visiting tuts.example.com just downloads the HTML.

I would like that any file (or preferably specific files) at that location would get that header applied. The folder can have pdf/txt/mp4 files (amongst others), and some will open in the browser and others will download, again depending on the browser used.

The only 100% reliable way with NGINX that I could have a simple indexed page _and_ force certain filetypes to download was to use the add_header method I mentioned above. I also tried changing the mime type to application/octet-stream but that wasn't reliable either.

Is this possible with Caddy or have I just messed up the syntax?

[mholt]

With the headers directive, you've set the base path to / which equates to all requests. It should instead be the path to only the folder which has the downloads stored in it.

[gavin19]

That's what I had initially but it didn't seem to work at all. I tried just adding a custom header with

header /var/www/static/tuts X-Custom-Header "Some value"

but I couldn't see that in the response in the Chrome dev tools.

If I test a file that I know will download by default, like a zip, then I'll get something like

Accept-Ranges:bytes
Connection:keep-alive
Content-Disposition:attachment
Content-Length:7
Content-Type:application/zip
Date:Mon, 18 Jan 2016 12:23:03 GMT
ETag:"569cd4c7-7"
Last-Modified:Mon, 18 Jan 2016 12:04:23 GMT
Server:nginx/1.4.6 (Ubuntu)

with NGINX, but with Caddy

accept-ranges:bytes
content-length:7
content-type:application/zip
date:Mon, 18 Jan 2016 12:18:05 GMT
last-modified:Mon, 18 Jan 2016 12:04:23 GMT
server:Caddy
status:200

The server blocks are identical between NGINX/Caddy. All I did was change the syntax where needed, like browse instead of autoindex on.

[Xe]

If your root is /var/www/static/, you should set the header for /tuts. It is doing matching based on the HTTP GET path.

[gavin19]

The root is /var/www/static/tuts. It's a folder with a bunch of files in it. When someone hits tuts.example.com then I want to serve the contents as a basic indexed page. This much works fine, barring the headers. I have it stripped down to simply

tuts.example.com {
    root /var/www/static/tuts
    browse
    header /var/www/static/tuts Content-disposition "attachment"
}

but no headers get applied.

For testing purposes, I tried making a files folder within tuts and moving the files in there, then used

tuts.example.com {
    root /var/www/static/tuts
    browse
    header /var/www/static/tuts/files Content-disposition "attachment"
}

but no difference.

[abiosoft]

header is request path based (not OS path based). Maybe try adding mime type too.

tuts.example.com {
    root /var/www/static/tuts
    browse
    header / Content-disposition "attachment"
    mime .mp4 application/octect-stream
    mime .pdf application/octect-stream
}

[gavin19]

That was my initial assumption, but when I implement it exactly as you mentioned then visiting tuts.example.com downloads the HTML, it doesn't render the index.

[abiosoft]

Oh, you mean it downloads every file, including html. But you only want it limited to mp4, pdf e.t.c ?

[gavin19]

No, I mean it literally downloads a copy of the HTML. If I go to that address it'll download download.htm. It doesn't actually take me to tuts.example.com.

[mholt]

That's what I've been saying -- the path you set for your header directive should only be the path for files you WANT the browser to download (in your case) -- not the path that includes the html files.

[gavin19]

There are no HTML files in that tuts folder. It's just a file dump for txt/zip/mp4/pdfs mostly. That's why I used browse so it would just display as a list of files to choose from to download.

If I use

tuts.example.com {
    root /var/www/static/tuts
    browse
    header /var/www/static/tuts Content-disposition "attachment"
}

then the page renders properly but none of the files get that header applied. If I use

tuts.example.com {
    root /var/www/static/tuts
    browse
    header / Content-disposition "attachment"
}

then it's as I described. The browser just downloads the HTML (i.e the HTML that would be used if the page was to be displayed).

[mholt]

Ohh. So what's happening is you're setting the header for all requests, which tells the browser to download a file no matter what the response is. The browse middleware just writes an HTML response - the browser doesn't know it's not a real file on disk, it just thinks it has to download it because you're setting that header.

[gavin19]

That's the problem. Either I get the page as a download, or the page renders as an indexed page but none of the downloadable files get that header applied.

With NGINX my setup was this

server {
    listen   80;

    root /var/www/static/tuts;
    server_name tuts.example.com;

    location ~* \.pdf$ {
        add_header Content-Disposition "attachment";
    }
    autoindex on;
}

Visiting the page would show a list of the files in the tuts folder, much like browse. The crucial difference being that it allowed me to target files inside the folder and apply a header only to them.

If I didn't specify a file(s) and just used

server {
    listen   80;

    root /var/www/static/tuts;
    server_name tuts.example.com;

    add_header Content-Disposition "attachment";

    autoindex on;
}

then it behaves exactly as

tuts.example.com {
    root /var/www/static/tuts
    browse
    header / Content-disposition "attachment"
}

in that it returns an HTML download, not a page listing.

So, as it stands, I can't use browse _and_ target files in that folder to apply headers to? Something like

header /*.pdf Content-disposition "attachment"

Thanks for the help btw. I was able to get as much done in 10 minutes with Caddy as it took me the best part of an hour with NGINX.

[mholt]

Ahhh, I see, so you need to apply headers to just .pdf files.

This can't be done right now -- not intuitively, anyway -- except maybe using a hack with the rewrite directive (@abiosoft?).

Meanwhile I'm considering implementing more powerful request matching features since the number of valid use cases seem to be adding up.

[abiosoft]

Actually, rewrite can as well be able to add/remove headers since proxy can do that.

[mholt]

Closing since the answer to the question is the same feature request as in a few other related issues. Thanks!

[squee1945]

Closing since the answer to the question is the same feature request as in a few other related issues.

I am trying to track down the status of this feature request. Specifically, I want to add headers to a rewrite rule - i.e., if I match a rewrite regex, I want to add response headers.

Is this possible today?